Skip to content

Varying noise graphs for 01/02 comparison

Diarmuid Coleman requested to merge noise-graphs-for-comparison-work into main

Adds slightly improved generic graphing functionality and addresses some points from !12 (merged)

The performance for lr = 0.01 remains very good for low to medium levels of noise, regardless of batch size, which would indicate that the differing results are most likely some form of precision inequality between the NumPy and PyTorch implementations.

Merge request reports